Sounds a bit on the small side, I'd use 20 at least, and maybe 30 to be on the safe side. This is not dependent on the k-point sampling, but does depend on the basis set used, plus you need to have enough vacuum to screen the electrostatic effects. Notably, in ATK vacuum is not "expensive" (as it is in a plane-wave code), provided you use the FFT or FFT2D solvers.
